PhD Final Defense – Zahra Heydari
Data-driven Approaches for Residential Water End-use Classification and Sustainable Urban Water Management
Advisor- Professor Ashlynn Stillwell
Abstract
Accurately measuring household water demand remains a major challenge for sustainable water management. Traditional methods such as surveys or aggregated billing data often lack the resolution and objectivity needed to inform targeted conservation or infrastructure strategies. As urban areas face growing water stress, there is increasing demand for approaches that provide detailed, reliable insights into how water is used within homes. This dissertation addresses that need by leveraging smart water systems and data-driven methods to non-intrusively study residential water use at the end-use level.
Using high-resolution data from smart water meters, this work explores how temporal granularity affects the ability to classify individual water-use events, highlighting the trade-offs between data precision, device limitations, and modeling performance. To expand the availability of training data, synthetic datasets are generated and used to evaluate a variety of machine learning models suited to the unique patterns of residential water use. The framework also introduces stagnation time, the period during which a fixture remains unused, as a novel metric for understanding behavior, detecting leaks, and assessing plumbing-related water quality risks. Together, these components form a comprehensive analytical approach that supports accurate disaggregation, robust model selection, and meaningful interpretation of water-use behavior.
By uncovering fine-grained insights into how water is used across fixtures and over time, smart water systems enable utilities to design more responsive infrastructure, identify inefficiencies, and guide policy decisions. For residents, the same insights can inform conservation habits, detect leaks early, and improve awareness of water-use behavior. This dissertation illustrates how combining advanced data analytics with smart metering technology can lead to more efficient and sustainable urban water management.